diff --git a/lib/Target/X86/X86FastISel.cpp b/lib/Target/X86/X86FastISel.cpp
index 08ad7fa11d7..4eeefae6318 100644
--- a/lib/Target/X86/X86FastISel.cpp
+++ b/lib/Target/X86/X86FastISel.cpp
@@ -1527,6 +1527,8 @@ static unsigned computeBytesPopedByCalle(const X86Subtarget &Subtarget,
     return 0;
   if (!CS.paramHasAttr(1, Attribute::StructRet))
     return 0;
+ if (CS.paramHasAttr(1, Attribute::InReg))
+    return 0;
   return 4;
 }
 
diff --git a/lib/Target/X86/X86ISelLowering.cpp b/lib/Target/X86/X86ISelLowering.cpp
index 4ccb0a391eb..40283d87f87 100644
--- a/lib/Target/X86/X86ISelLowering.cpp
+++ b/lib/Target/X86/X86ISelLowering.cpp
@@ -1717,21 +1717,37 @@ X86TargetLowering::LowerCallResult(SDValue Chain, SDValue InFlag,
 
 /// CallIsStructReturn - Determines whether a call uses struct return
 /// semantics.
-static bool CallIsStructReturn(const SmallVectorImpl<ISD::OutputArg> &Outs) {
+enum StructReturnType {
+  NotStructReturn,
+  RegStructReturn,
+  StackStructReturn
+};
+static StructReturnType
+callIsStructReturn(const SmallVectorImpl<ISD::OutputArg> &Outs) {
   if (Outs.empty())
-    return false;
+    return NotStructReturn;
 
-  return Outs[0].Flags.isSRet();
+  const ISD::ArgFlagsTy &Flags = Outs[0].Flags;
+  if (!Flags.isSRet())
+    return NotStructReturn;
+  if (Flags.isInReg())
+    return RegStructReturn;
+  return StackStructReturn;
 }
 
 /// ArgsAreStructReturn - Determines whether a function uses struct
 /// return semantics.
-static bool
-ArgsAreStructReturn(const SmallVectorImpl<ISD::InputArg> &Ins) {
+static StructReturnType
+argsAreStructReturn(const SmallVectorImpl<ISD::InputArg> &Ins) {
   if (Ins.empty())
-    return false;
+    return NotStructReturn;
 
-  return Ins[0].Flags.isSRet();
+  const ISD::ArgFlagsTy &Flags = Ins[0].Flags;
+  if (!Flags.isSRet())
+    return NotStructReturn;
+  if (Flags.isInReg())
+    return RegStructReturn;
+  return StackStructReturn;
 }
 
 /// CreateCopyOfByValArgument - Make a copy of an aggregate at address specified
@@ -2072,7 +2088,7 @@ X86TargetLowering::LowerFormalArguments(SDValue Chain,
     FuncInfo->setBytesToPopOnReturn(0); // Callee pops nothing.
     // If this is an sret function, the return should pop the hidden pointer.
     if (!Is64Bit && !IsTailCallConvention(CallConv) && !IsWindows &&
-        ArgsAreStructReturn(Ins))
+        argsAreStructReturn(Ins) == StackStructReturn)
       FuncInfo->setBytesToPopOnReturn(4);
   }
 
@@ -2162,7 +2178,7 @@ X86TargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
   bool Is64Bit        = Subtarget->is64Bit();
   bool IsWin64        = Subtarget->isTargetWin64();
   bool IsWindows      = Subtarget->isTargetWindows();
-  bool IsStructRet    = CallIsStructReturn(Outs);
+  StructReturnType SR = callIsStructReturn(Outs);
   bool IsSibcall      = false;
 
   if (MF.getTarget().Options.DisableTailCalls)
@@ -2171,8 +2187,9 @@ X86TargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
   if (isTailCall) {
     // Check if it's really possible to do a tail call.
     isTailCall = IsEligibleForTailCallOptimization(Callee, CallConv,
-                    isVarArg, IsStructRet, MF.getFunction()->hasStructRetAttr(),
-                                                   Outs, OutVals, Ins, DAG);
+                    isVarArg, SR != NotStructReturn,
+                    MF.getFunction()->hasStructRetAttr(),
+                    Outs, OutVals, Ins, DAG);
 
     // Sibcalls are automatically detected tailcalls which do not require
     // ABI changes.
@@ -2548,7 +2565,7 @@ X86TargetLowering::LowerCall(TargetLowering::CallLoweringInfo &CLI,
                        getTargetMachine().Options.GuaranteedTailCallOpt))
     NumBytesForCalleeToPush = NumBytes;    // Callee pops everything
   else if (!Is64Bit && !IsTailCallConvention(CallConv) && !IsWindows &&
-           IsStructRet)
+           SR == StackStructReturn)
     // If this is a call to a struct-return function, the callee
     // pops the hidden struct pointer, so we have to push it back.
     // This is common for Darwin/X86, Linux & Mingw32 targets.
diff --git a/test/CodeGen/X86/inreg.ll b/test/CodeGen/X86/inreg.ll
index a6a7ebada38..6653cfb14ed 100644
--- a/test/CodeGen/X86/inreg.ll
+++ b/test/CodeGen/X86/inreg.ll
@@ -1,19 +1,46 @@
-; RUN: llc < %s -mtriple=i686-pc-linux -mcpu=corei7 | FileCheck %s
+; RUN: llc < %s -mtriple=i686-pc-linux -mcpu=corei7 | FileCheck --check-prefix=DAG %s
+; RUN: llc < %s -mtriple=i686-pc-linux -mcpu=corei7 -O0 | FileCheck --check-prefix=FAST %s
 
-%struct.s = type { double, float }
+%struct.s1 = type { double, float }
 
-define void @g() nounwind {
+define void @g1() nounwind {
 entry:
-  %tmp = alloca %struct.s, align 4
-  call void @f(%struct.s* inreg sret %tmp, i32 inreg 41, i32 inreg 42, i32 43)
+  %tmp = alloca %struct.s1, align 4
+  call void @f(%struct.s1* inreg sret %tmp, i32 inreg 41, i32 inreg 42, i32 43)
   ret void
-  ; CHECK: g:
-  ; CHECK: subl {{.*}}, %esp
-  ; CHECK-NEXT: $43, (%esp)
-  ; CHECK-NEXT: leal    16(%esp), %eax
-  ; CHECK-NEXT: movl    $41, %edx
-  ; CHECK-NEXT: movl    $42, %ecx
-  ; CHECK-NEXT: calll   f
+  ; DAG: g1:
+  ; DAG: subl $[[AMT:.*]], %esp
+  ; DAG-NEXT: $43, (%esp)
+  ; DAG-NEXT: leal    16(%esp), %eax
+  ; DAG-NEXT: movl    $41, %edx
+  ; DAG-NEXT: movl    $42, %ecx
+  ; DAG-NEXT: calll   f
+  ; DAG-NEXT: addl $[[AMT]], %esp
+  ; DAG-NEXT: ret
+
+  ; FAST: g1:
+  ; FAST: subl $[[AMT:.*]], %esp
+  ; FAST-NEXT: leal    8(%esp), %eax
+  ; FAST-NEXT: movl    $41, %edx
+  ; FAST-NEXT: movl    $42, %ecx
+  ; FAST: $43, (%esp)
+  ; FAST: calll   f
+  ; FAST-NEXT: addl $[[AMT]], %esp
+  ; FAST: ret
 }
 
-declare void @f(%struct.s* inreg sret, i32 inreg, i32 inreg, i32)
+declare void @f(%struct.s1* inreg sret, i32 inreg, i32 inreg, i32)
+
+%struct.s2 = type {}
+
+define void @g2(%struct.s2* inreg sret %agg.result) nounwind {
+entry:
+  ret void
+  ; DAG: g2
+  ; DAG-NOT: ret $4
+  ; DAG: .size g2
+
+  ; FAST: g2
+  ; FAST-NOT: ret $4
+  ; FAST: .size g2
+}